vfio: simplify iommu group allocation for mediated devices

Reuse the logic in vfio_noiommu_group_alloc to allocate a fake
single-device iommu group for mediated devices by factoring out a common
function, and replacing the noiommu boolean field in struct vfio_group
with an enum to distinguish the three different kinds of groups.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Kevin Tian <kevin.tian@intel.com>
Link: https://lore.kernel.org/r/20210924155705.4258-8-hch@lst.de
Signed-off-by: Alex Williamson <alex.williamson@redhat.com>

1 files changed, 66 insertions, 26 deletions

diff --git a/drivers/vfio/vfio.c b/drivers/vfio/vfio.c
index 23eaebd2e28c..2508c8c39840 100644
--- a/drivers/vfio/vfio.c
+++ b/drivers/vfio/vfio.c
@@ -67,6 +67,30 @@ struct vfio_unbound_dev {
 	struct list_head		unbound_next;
 };
 
+enum vfio_group_type {
+	/*
+	 * Physical device with IOMMU backing.
+	 */
+	VFIO_IOMMU,
+
+	/*
+	 * Virtual device without IOMMU backing. The VFIO core fakes up an
+	 * iommu_group as the iommu_group sysfs interface is part of the
+	 * userspace ABI.  The user of these devices must not be able to
+	 * directly trigger unmediated DMA.
+	 */
+	VFIO_EMULATED_IOMMU,
+
+	/*
+	 * Physical device without IOMMU backing. The VFIO core fakes up an
+	 * iommu_group as the iommu_group sysfs interface is part of the
+	 * userspace ABI.  Users can trigger unmediated DMA by the device,
+	 * usage is highly dangerous, requires an explicit opt-in and will
+	 * taint the kernel.
+	 */
+	VFIO_NO_IOMMU,
+};
+
 struct vfio_group {
 	struct kref			kref;
 	int				minor;
@@ -83,7 +107,7 @@ struct vfio_group {
 	struct mutex			unbound_lock;
 	atomic_t			opened;
 	wait_queue_head_t		container_q;
-	bool				noiommu;
+	enum vfio_group_type		type;
 	unsigned int			dev_counter;
 	struct kvm			*kvm;
 	struct blocking_notifier_head	notifier;
@@ -336,7 +360,7 @@ static void vfio_group_unlock_and_free(struct vfio_group *group)
  * Group objects - create, release, get, put, search
  */
 static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group,
-		bool noiommu)
+		enum vfio_group_type type)
 {
 	struct vfio_group *group, *tmp;
 	struct device *dev;
@@ -355,7 +379,7 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group,
 	atomic_set(&group->opened, 0);
 	init_waitqueue_head(&group->container_q);
 	group->iommu_group = iommu_group;
-	group->noiommu = noiommu;
+	group->type = type;
 	BLOCKING_INIT_NOTIFIER_HEAD(&group->notifier);
 
 	group->nb.notifier_call = vfio_iommu_group_notifier;
@@ -391,8 +415,8 @@ static struct vfio_group *vfio_create_group(struct iommu_group *iommu_group,
 	}
 
 	dev = device_create(vfio.class, NULL,
-			    MKDEV(MAJOR(vfio.group_devt), minor),
-			    group, "%s%d", group->noiommu ? "noiommu-" : "",
+			    MKDEV(MAJOR(vfio.group_devt), minor), group, "%s%d",
+			    group->type == VFIO_NO_IOMMU ? "noiommu-" : "",
 			    iommu_group_id(iommu_group));
 	if (IS_ERR(dev)) {
 		vfio_free_group_minor(minor);
@@ -778,8 +802,8 @@ void vfio_uninit_group_dev(struct vfio_device *device)
 }
 EXPORT_SYMBOL_GPL(vfio_uninit_group_dev);
 
-#ifdef CONFIG_VFIO_NOIOMMU
-static struct vfio_group *vfio_noiommu_group_alloc(struct device *dev)
+static struct vfio_group *vfio_noiommu_group_alloc(struct device *dev,
+		enum vfio_group_type type)
 {
 	struct iommu_group *iommu_group;
 	struct vfio_group *group;
@@ -794,7 +818,7 @@ static struct vfio_group *vfio_noiommu_group_alloc(struct device *dev)
 	if (ret)
 		goto out_put_group;
 
-	group = vfio_create_group(iommu_group, true);
+	group = vfio_create_group(iommu_group, type);
 	if (IS_ERR(group)) {
 		ret = PTR_ERR(group);
 		goto out_remove_device;
@@ -808,7 +832,6 @@ out_put_group:
 	iommu_group_put(iommu_group);
 	return ERR_PTR(ret);
 }
-#endif
 
 static struct vfio_group *vfio_group_find_or_alloc(struct device *dev)
 {
@@ -824,7 +847,7 @@ static struct vfio_group *vfio_group_find_or_alloc(struct device *dev)
 		 * bus.  Taint the kernel because we're about to give a DMA
 		 * capable device to a user without IOMMU protection.
 		 */
-		group = vfio_noiommu_group_alloc(dev);
+		group = vfio_noiommu_group_alloc(dev, VFIO_NO_IOMMU);
 		if (!IS_ERR(group)) {
 			add_taint(TAINT_USER, LOCKDEP_STILL_OK);
 			dev_warn(dev, "Adding kernel taint for vfio-noiommu group on device\n");
@@ -841,7 +864,7 @@ static struct vfio_group *vfio_group_find_or_alloc(struct device *dev)
 		goto out_put;
 
 	/* a newly created vfio_group keeps the reference. */
-	group = vfio_create_group(iommu_group, false);
+	group = vfio_create_group(iommu_group, VFIO_IOMMU);
 	if (IS_ERR(group))
 		goto out_put;
 	return group;
@@ -851,10 +874,13 @@ out_put:
 	return group;
 }
 
-int vfio_register_group_dev(struct vfio_device *device)
+static int __vfio_register_dev(struct vfio_device *device,
+		struct vfio_group *group)
 {
 	struct vfio_device *existing_device;
-	struct vfio_group *group;
+
+	if (IS_ERR(group))
+		return PTR_ERR(group);
 
 	/*
 	 * If the driver doesn't specify a set then the device is added to a
@@ -863,16 +889,13 @@ int vfio_register_group_dev(struct vfio_device *device)
 	if (!device->dev_set)
 		vfio_assign_device_set(device, device);
 
-	group = vfio_group_find_or_alloc(device->dev);
-	if (IS_ERR(group))
-		return PTR_ERR(group);
-
 	existing_device = vfio_group_get_device(group, device->dev);
 	if (existing_device) {
 		dev_WARN(device->dev, "Device already exists on group %d\n",
 			 iommu_group_id(group->iommu_group));
 		vfio_device_put(existing_device);
-		if (group->noiommu)
+		if (group->type == VFIO_NO_IOMMU ||
+		    group->type == VFIO_EMULATED_IOMMU)
 			iommu_group_remove_device(device->dev);
 		vfio_group_put(group);
 		return -EBUSY;
@@ -891,8 +914,25 @@ int vfio_register_group_dev(struct vfio_device *device)
 
 	return 0;
 }
+
+int vfio_register_group_dev(struct vfio_device *device)
+{
+	return __vfio_register_dev(device,
+		vfio_group_find_or_alloc(device->dev));
+}
 EXPORT_SYMBOL_GPL(vfio_register_group_dev);
 
+/*
+ * Register a virtual device without IOMMU backing.  The user of this
+ * device must not be able to directly trigger unmediated DMA.
+ */
+int vfio_register_emulated_iommu_dev(struct vfio_device *device)
+{
+	return __vfio_register_dev(device,
+		vfio_noiommu_group_alloc(device->dev, VFIO_EMULATED_IOMMU));
+}
+EXPORT_SYMBOL_GPL(vfio_register_emulated_iommu_dev);
+
 /**
  * Get a reference to the vfio_device for a device.  Even if the
  * caller thinks they own the device, they could be racing with a
@@ -1019,7 +1059,7 @@ void vfio_unregister_group_dev(struct vfio_device *device)
 	if (list_empty(&group->device_list))
 		wait_event(group->container_q, !group->container);
 
-	if (group->noiommu)
+	if (group->type == VFIO_NO_IOMMU || group->type == VFIO_EMULATED_IOMMU)
 		iommu_group_remove_device(device->dev);
 
 	/* Matches the get in vfio_register_group_dev() */
@@ -1368,7 +1408,7 @@ static int vfio_group_set_container(struct vfio_group *group, int container_fd)
 	if (atomic_read(&group->container_users))
 		return -EINVAL;
 
-	if (group->noiommu && !capable(CAP_SYS_RAWIO))
+	if (group->type == VFIO_NO_IOMMU && !capable(CAP_SYS_RAWIO))
 		return -EPERM;
 
 	f = fdget(container_fd);
@@ -1388,7 +1428,7 @@ static int vfio_group_set_container(struct vfio_group *group, int container_fd)
 
 	/* Real groups and fake groups cannot mix */
 	if (!list_empty(&container->group_list) &&
-	    container->noiommu != group->noiommu) {
+	    container->noiommu != (group->type == VFIO_NO_IOMMU)) {
 		ret = -EPERM;
 		goto unlock_out;
 	}
@@ -1402,7 +1442,7 @@ static int vfio_group_set_container(struct vfio_group *group, int container_fd)
 	}
 
 	group->container = container;
-	container->noiommu = group->noiommu;
+	container->noiommu = (group->type == VFIO_NO_IOMMU);
 	list_add(&group->container_next, &container->group_list);
 
 	/* Get a reference on the container and mark a user within the group */
@@ -1426,7 +1466,7 @@ static int vfio_group_add_container_user(struct vfio_group *group)
 	if (!atomic_inc_not_zero(&group->container_users))
 		return -EINVAL;
 
-	if (group->noiommu) {
+	if (group->type == VFIO_NO_IOMMU) {
 		atomic_dec(&group->container_users);
 		return -EPERM;
 	}
@@ -1451,7 +1491,7 @@ static int vfio_group_get_device_fd(struct vfio_group *group, char *buf)
 	    !group->container->iommu_driver || !vfio_group_viable(group))
 		return -EINVAL;
 
-	if (group->noiommu && !capable(CAP_SYS_RAWIO))
+	if (group->type == VFIO_NO_IOMMU && !capable(CAP_SYS_RAWIO))
 		return -EPERM;
 
 	device = vfio_device_get_from_name(group, buf);
@@ -1498,7 +1538,7 @@ static int vfio_group_get_device_fd(struct vfio_group *group, char *buf)
 
 	fd_install(fdno, filep);
 
-	if (group->noiommu)
+	if (group->type == VFIO_NO_IOMMU)
 		dev_warn(device->dev, "vfio-noiommu device opened by user "
 			 "(%s:%d)\n", current->comm, task_pid_nr(current));
 	return fdno;
@@ -1594,7 +1634,7 @@ static int vfio_group_fops_open(struct inode *inode, struct file *filep)
 	if (!group)
 		return -ENODEV;
 
-	if (group->noiommu && !capable(CAP_SYS_RAWIO)) {
+	if (group->type == VFIO_NO_IOMMU && !capable(CAP_SYS_RAWIO)) {
 		vfio_group_put(group);
 		return -EPERM;
 	}